Junping Du commented on YARN-3334:

Thanks [~zjshen] and [~sjlee0] for comments!
bq. If so, I suggest combining the two massages together, and record a 
error-level log (the first message is actually useless, if we always report the 
second one).
That sounds OK. Will update a quick fix.

bq. However, I do worry about the size of the map produced in the response in 
ResourceTrackerService. It can be potentially quite large every time and has a 
potential impact on many things as it is part of the NM heartbeat handling. 
It's OK for now, but we should try to address it sooner than later.
Just filed YARN-3445 to track this issue. This is also needed in gracefully 
decommission (YARN-914) - decommissioning node can be terminated earlier by RM 
if no running apps.

> [Event Producers] NM TimelineClient life cycle handling and container metrics 
> posting to new timeline service.
> --------------------------------------------------------------------------------------------------------------
>                 Key: YARN-3334
>                 URL: https://issues.apache.org/jira/browse/YARN-3334
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager
>    Affects Versions: YARN-2928
>            Reporter: Junping Du
>            Assignee: Junping Du
>         Attachments: YARN-3334-demo.patch, YARN-3334-v1.patch, 
> YARN-3334-v2.patch, YARN-3334-v3.patch, YARN-3334-v4.patch, 
> YARN-3334-v5.patch, YARN-3334-v6.patch, YARN-3334.7.patch
> After YARN-3039, we have service discovery mechanism to pass app-collector 
> service address among collectors, NMs and RM. In this JIRA, we will handle 
> service address setting for TimelineClients in NodeManager, and put container 
> metrics to the backend storage.

This message was sent by Atlassian JIRA

Reply via email to